Nitric Oxide Synthase Inhibitors That Interact with Both Heme

Article pubs.acs.org/jmc

Open Access on 04/23/2015

Nitric Oxide Synthase Inhibitors That Interact with Both Heme Propionate and Tetrahydrobiopterin Show High Isoform Selectivity Soosung Kang,† Wei Tang,† Huiying Li,‡ Georges Chreifi,‡ Pavel Martásek,§ Linda J. Roman,§ Thomas L. Poulos,*,‡ and Richard B. Silverman*,† †

Department of Chemistry, Department of Molecular Biosciences, Chemistry of Life Processes Institute, Center for Molecular Innovation and Drug Discovery, Northwestern University, Evanston, Illinois 60208-3113, United States ‡ Departments of Molecular Biology and Biochemistry, Pharmaceutical Sciences, and Chemistry, University of California, Irvine, California 92697-3900, United States § Department of Biochemistry, University of Texas Health Science Center, San Antonio, Texas 78384-7760, United States S Supporting Information *

ABSTRACT: Overproduction of NO by nNOS is implicated in the pathogenesis of diverse neuronal disorders. Since NO signaling is involved in diverse physiological functions, selective inhibition of nNOS over other isoforms is essential to minimize side effects. A series of α-amino functionalized aminopyridine derivatives (3−8) were designed to probe the structure−activity relationship between ligand, heme propionate, and H4B. Compound 8R was identified as the most potent and selective molecule of this study, exhibiting a Ki of 24 nM for nNOS, with 273-fold and 2822-fold selectivity against iNOS and eNOS, respectively. Although crystal structures of 8R complexed with nNOS and eNOS revealed a similar binding mode, the selectivity stems from the distinct electrostatic environments in two isoforms that result in much lower inhibitor binding free energy in nNOS than in eNOS. These findings provide a basis for further development of simple, but even more selective and potent, nNOS inhibitors.

■

various cancers.13 In addition, NO synthesis from bacterial NOS was reported to play a critical role in antibiotic resistance and pathogenicity.14,15 This suggests that the inhibition of NOSs can be effective for the control of these diverse diseases, but because NO signaling is involved in various physiological functions, selective inhibition is essential to minimize any unwanted side effects.16 NOSs are homodimeric enzymes; each monomer consists of a reductase domain and an oxygenase domain. A C-terminal reductase domain contains NADPH, FAD, and FMN cofactors,17 and an N-terminal oxygenase domain contains iron protoporphyrin IX (heme), where the substrate L-Arg binds, and tetrahydrobiopterin (H4B) cofactors.18 H4B forms tight H-bonds with the propionate of the heme A-ring and provides an electron that is crucial for activating the hemebound dioxygen during the catalytic reaction.19 Although H4B

INTRODUCTION The free radical nitric oxide (NO) is an important signaling molecule,1 controlling diverse physiological and pathological processes in various species.2 In mammals, NO is endogenously produced using L-arginine and molecular oxygen with NADPH by three principal nitric oxide synthases (NOSs): neuronal NOS (nNOS), endothelial NOS (eNOS), and inducible NOS (iNOS).3 Selective inhibition of each NOS can regulate different biological functions of NO signaling because each NOS isoform is localized differently in the neuron, endothelium, and immune system, and is activated by a specific pathway.4 Overproduction of NO by nNOS in the central nervous system has been implicated in the pathogenesis of diverse neuronal disorders such as strokes,5 septic shock,6 seizures,7 migraine headaches,8 Alzheimer’s disease,9 Parkinson’s disease,10 and ALS.11 Recently, nNOS has also been implicated to play a critical role in melanoma tumor development and growth.12 In the immune system, excess NO production from iNOS is also linked to inflammation and © 2014 American Chemical Society

Received: March 17, 2014 Published: April 23, 2014 4382

dx.doi.org/10.1021/jm5004182 | J. Med. Chem. 2014, 57, 4382−4396

Journal of Medicinal Chemistry

Article

Figure 1. (a) Chemical structures of leads 1 and 2 and inhibitory activities; (b) overlay of inhibitors 1 (cyan) and 2 (orange) complexed with nNOS, showing the heme (pink), H4B (yellow), and key residues (green) in the active site (PDB: 3NLM and 3N5W); and (c) the proposed scaffold derived from compounds 1 and 2.

Figure 2. Prepared and tested molecules in this study.

and the heme A-ring propionate, replacing a water molecule, while the fluorophenyl ring stacks with the heme plane. Despite the excellent isoform selectivity of this molecule, the construction of the two unnaturally occurring chiral centers of 1 is not efficient and requires multiple steps with a low overall yield. This limits the opportunities for optimizing the pharmacokinetic properties of the inhibitor and for carrying out in vivo studies. Compound 2, the other potent nNOS inhibitor (Ki = 25 nM), is only moderately selective (i/n = 58, e/n = 107) but can be prepared from commercial starting materials in four chemical steps, in an excellent overall yield. The common feature of 1 and 2 is that both utilize one aminopyridine to make H-bonds with the heme D-ring propionate and to stack with Tyr706 (Figure 1B). The additional H-bonds between the other aminopyridine of 2 and Glu592 of nNOS anchor the inhibitor to the substrate binding site above the heme in a double-headed mode. However, 2, in contrast to 1, does not

binding is not required for dimerization, it interacts with both subunits of the dimer by forming part of the dimerization interface to enrich the structural stability of the dimer.20,21 For over a decade, our research groups have been interested in the development of selective inhibitors of nNOS for the treatment of neurodegenerative disease. Among diverse NOS inhibitors, compounds 122 and 223 (Figure 1A) are the most potent inhibitors for nNOS. They are spotlighted by excellent isoform selectivity for 1 and easy synthesis for 2. Compound 1 has >700-fold selectivity against iNOS and >3800-fold selectivity against eNOS. The X-ray crystal structures of 1 complexed with nNOS and eNOS24 reveal features of enzyme− inhibitor interactions that form the basis for high potency and selectivity (Figure 1B): the aminopyridine of 1 interacts with a heme D-ring propionate via two H-bonds, as well as with Tyr706 in a π−π stacking interaction. The pyrrolidine nitrogen of 1 is located within hydrogen-bonding distances to both H4B 4383



Article

Scheme 1. Synthesis of 3R and 3Sa

Reagents and conditions: (a) 9 (2.5 equiv), BuLi (2.5 equiv), THF, 0 °C to −78 °C; (b) DEAD, DPPA, PPh3, THF, room temp., 12 h; (c) LiAlH4; (d) (i) (S)-camphamic chloride, TEA, CH2Cl2, room temp., (ii) chiral resolution on a silica gel column; (e) conc HCl, AcOH, microwave, 150 °C, 5 h.

a

Scheme 2. Synthesis of 5R and 5Sa

a Reagents and conditions; (a) BuLi, THF; (b) DIBAL; (c) (S)-t-butylsulfinamide, Ti(OEt)4; (d) 9, BuLi; (e) conc HCl, EtOH, microwave, 120 °C, 20 min.

directly interact with the H4B or with the propionate of the heme A-ring. The lack of these interactions may explain the moderate selectivity of 2 against iNOS and eNOS. A possible strategy to confirm this structure-selectivity issue is to install a new functional group on molecule 2 to create an interaction with H4B and the heme A-ring propionate. This approach will allow us to perform a SAR study in this area for a

new chemotype design in which the molecules will be easy to prepare while still being highly isoform selective. From the structure overlay of 1 and 2 (Figure 1B), we found that the pyrrolidine amine of 1 could be replaced by the addition of an amino group on the side chain of 2 (Figure 1C). The alignment of an amino group with an adapting position of the middle aromatic ring of 2, especially maintaining the structural 4384



Article

Scheme 3. Synthesis of 5a

a

Reagents and conditions: (a) (i) 9, BuLi, THF; (ii) 15; (b) DIBAL-H; (c) (i) MeNO2, TEA, (ii) AcCl; (d) 9, BuLi; (e) LAH; (f) conc HCl, EtOH, microwave, 120 °C, 20 min.

Scheme 4. Synthesis of 7 and 8a

a Reagents and conditions: (a) (i) MeNO2, TEA, (ii) AcCl; (b) 9, BuLi; (c) 28, Pd(PPh3)2Cl2, CuI, PPh3, DEA, DMF; (d) Raney-Ni, H2, MeOH/ EtOH; (e) conc HCl/EtOH (1/2), microwave, 120 °C, 20 min.

■

CHEMISTRY The synthesis of compounds 3R and 3S is shown in Scheme 1. Benzyl alcohol 11 was prepared by coupling of 3bromomethylbenzaldehyde (9) with two equivalents of lithiated pyrrolyl-4,6-lutidine (10). The hydroxyl group of 11 was then converted to benzyl azide 12 via a Mitsunobu reaction with DPPA. Reduction of the azide with LiAlH4 gave the free amine, which subsequently underwent amidation with (S)camphanic chloride to give a separable diastereomeric mixture. Each pure diastereomer, 13a and 13b, was successfully isolated using general silica gel column chromatography. An asymmetric approach toward the target compounds using Ellman’s chiral sulfinamide in the synthesis of 4R and 4S (Scheme 2) was not successful; only inseparable diastereomeric mixtures were produced. The (S)-camphanyl auxiliary and the two protecting groups on the aminopyridine rings were removed together by

similarity to the pyrrolidine amine of 1, was virtually performed using Surflex-Sim in the Sybyl-X program. Although the Renantiomer of the α-amino derivative was predicted to interact with H4B and a propionate of the heme, the preparation of the other enantiomer was also desirable to confirm the stereoactivity relationship. After 2-amino-4-methylpyridine was selected as a head near the α-amino group, three different meta-substituted aromatic rings were adapted as linkers, and another 2-amino-4-methylpyridine or a 4-methylpyridine ring was chosen as the second head in the hopes that it would stay above the heme with a proposed interaction with Glu592. The crystal structural information gathered with compounds bearing an α-amino group (3−5) led to the introduction of an aminomethyl group (6−8) to further improve the binding affinity and selectivity. Figure 2 summarizes the prepared and assayed compounds in this study. 4385



Article

Table 1. List of Ki Values and Selectivity of 3−8

a

Reference molecules that were reported previously.23

and 19b were removed by microwave-aided acidic hydrolysis to give 5R and 5S in high yields. Compound 6 was also prepared from benzyl bromide 15 using a six-step procedure (Scheme 3). Coupling of lithiated 9 with 15 gave nitrile 20, which was then reduced to aldehyde 21 by treatment with DIBAL. Condensation of 21 with nitromethane gave nitrovinyl compound 22 in good yield. Michael addition of 22 with lithiated 9 produced nitro intermediate 23, which was reduced to amine 24 using Raney-Ni under a hydrogen atmosphere. Both aminopyridine protecting groups were removed by microwave-aided hydrolysis to give 6 in good yields. Compounds 7 and 8 were prepared in five steps from commercially available brominated pyridinylaldehydes 25a−b

microwave-aided hydrolysis to give optically active 3R and 3S (Scheme 1). Compounds 5R and 5S were prepared from 2,4-dimethyllutidine and 15 using a five-step procedure (Scheme 2). Lithiated 2,4-dimethyllutidine was coupled with benzyl bromide 15 to give nitrile 16. The cyano group of 16 was reduced to an aldehyde (17) using DIBAL, which then underwent condensation with Ellman’s chiral sulfinamide to give (S)-N-tertbutanesulfinyl aldimine 18 in a moderate yield. This intermediate was coupled with lithiated pyrrolyl-4,6-lutidine 9 to give diastereomeric mixture 19a−b. Minor diasteromeric product 19b eluted first, and the major product 19a eluted second during silica gel column chromatography. Protecting groups on the aminopyridine and t-butyl sulfinamide of 19a 4386



Article

Table 2. Ki Values and Selectivity of 8S and 8R

Figure 3. Active site structures of rat nNOS in complex with (A) 3S (PDB: 4CTP), (B) 3R (PDB: 4CTQ), (C) 4 (PDB: 4CTR), and (D) 5 or 5S (PDB: 4CTT). The Fo − Fc omit electron density for the inhibitor is displayed at the 2.5 σ contour level. The weaker density in the tail part of 4 and 5 indicates disordering. The major H-bonds are depicted with dashed lines. All crystal structure figures were prepared with PyMol (www.pymol.org).

mandelic chloride also were not successful because the molecules have multiple reactive amines.

(Scheme 4). Condensation of 25a−b with nitromethane in the presence of TEA and acetyl chloride, followed by Michael reaction with lithiated 4,6-dimethylpyridine afforded 27a−b. Sonogashira coupling between 27a−b and alkyne 28, which was prepared from 6-bromo-2-aminopyridine and trimethylsilylacetylene, produced intermediate 29a−b. Raney nickelmediated hydrogenation of 29a−b yielded reduced product 30a−b. Finally, the protecting groups on the aminopyridine rings were removed using microwave conditions to yield products 7 and 8. A portion of intermediate 30b was protected with Boc and then injected onto an OD-H chiral HPLC column to separate the enantiomers. Each enantomerically pure compound was deprotected via microwave-assisted acidic hydrolysis to give optically active enantiomers 8S and 8R. Our diverse approach for the chiral resolution of the final deprotected compound (8) using CrownPak CR-(+), Chiralcel OD-RH, and Whelk-O 1 chiral reverse phase HPLC columns, and a Chiralcel OD-H chiral normal phase column with DEA, were not successful. Chiral derivatization of the final compound using Mosher’s acid chloride, (S)-camphanyl chloride, and (S)-

■

RESULTS AND DISCUSSION In Vitro Inhibitory Assays. The NOS isoform assays involved subjecting 3−8 to an oxyhemoglobin NO capture assay using a Biotek Gen5 microplate reader. IC50 values for each compound were determined in duplicate or triplicate using dose−response curves with nine concentration points (1 pM−3 mM). The standard deviation of the assays were less than 15% with nNOS or iNOS and less than 25% with eNOS. The inhibition constants (Ki) of these compounds were determined from the IC50 and Km values (rat nNOS = 1.3 μM; murine iNOS = 8.2 μM; and bovine eNOS = 1.7 μM) for all three NOS isoforms using the following relationship: Ki = IC50/(1 + [S]/KM) The selectivity of antagonism of nNOS relative to iNOS or eNOS was determined by calculating the ratios of the Ki values with iNOS or eNOS to those with nNOS. Compounds 3−8, having various amino functional groups, were found to have moderate to excellent selectivity (50−2822 of e/n, 36−273 of 4387



Article

hydroxyl group forms H-bonds with Glu592 so that the hydroxyl points toward the heme, similar to the position of the amino group in 3S, but the second aminopyridine is partially disordered, hanging in the active site access channel and making only van der Waals contacts with the protein. The structure of nNOS complexed with 5 was also checked to explore the binding mode for the compound bearing asymmetric double head groups. As expected, only the aminopyridine is able to anchor the inhibitor above the heme via H-bonds with Glu592, while the 4-methylpyridine head is poorly defined in the active site access channel (Figure 3D). Although a racemic mixture of 5S and 5R was used in crystal soaking, only the 5S enantiomer was picked up by the nNOS in the crystal, which agrees with the better affinity of 5S compared to that of 5R (Table 1). The α-amino nitrogen is also able to H-bond with Glu592, which forces the center phenyl ring to bend toward the heme, causing a distortion of the propionate of the heme D-ring. Both the α-amino group and the aminopyridine ring in 5S wrap around the Glu592 side chain, which resembles the binding mode of the substrate, L-Arg, where the α-amino and guanidinium groups embrace Glu592. Note that the α-amino group from the same S-chiral centers of 3S and 5S ends up with two totally different positions, H-bonding either with a water molecule (3S) or with the heme D-ring propionate (5S), because the binding orientation of 3S and 5S is flipped by 180° relative to each other. Both 3S and 5S showed modest isoform selectivity for nNOS over eNOS. We, therefore, also determined the crystal structures of 3S, 3R, and 5S bound to eNOS. In the eNOS3S structure (data not shown), only one aminopyridine was visible in the electron density that H-bonds with Glu363; the rest of 3S was disordered, which prevented us from obtaining a fully refined structure. Nevertheless, double-headed binding, as seen in the nNOS-3S structure, can be ruled out because the Tyr477 side chain still H-bonds with the propionate of the heme D-ring, thus blocking the access of the second aminopyridine to the site. The one-headed binding of 3S results in its poorer affinity to eNOS, leading to ∼500-fold selection for nNOS over eNOS (Table 1). In contrast, 3R binds to eNOS in a bouble-headed mode, as seen in Figure 4A, which is almost no different from what is observed in the nNOS-3R structure (Figure 3B), thus giving 5-fold improved affinity to eNOS compared with that of 3S (Table 1). Although having a similar binding mode to both nNOS and eNOS, 3R still shows 118-fold better affinity to nNOS than eNOS. Below, we discuss in more detail what gives rise to the isoform selection when the binding mode for an inhibitor is identical in the two isoforms. The structure of eNOS-5S shows some intriguing features (Figure 4B). While the aminopyridine makes H-bonds with Glu363, which is the same as that seen in the nNOS-5S structure, the α-amino group does not directly H-bond with Glu363. Instead, a water molecule is bridging in between. This binding preference resembles what we have observed for some nNOS selective dipeptide amide inhibitors, where the α-amino group of the dipeptide inhibitor made a direct H-bond with Glu592 in nNOS but was bridged by a water molecule in between the α-amino of the inhibitor and Glu363 in eNOS.26 This different binding preference results from the better electrostatic stabilization that the α-amino group of the inhibitor experiences by its proximity next to two negatively charged residues, Glu592 and Asp597 in nNOS, compared with just one negatively charged residue, Glu363 and Asn368 in eNOS. Therefore, the loss of the electrostatic interactions in

i/n) and moderate to good binding affinity (24−4370 nM) to nNOS. All of the synthesized molecules as well as three reference molecules (31, 32, and 2; for comparative purpose) and their associated activities are shown in Tables 1 and 2. The compounds with an α-amino (or a hydroxyl) group and symmetric double heads, 3S, 3R, and 4, exhibited about a 100 nM binding affinity to nNOS and modest selectivity against eNOS and iNOS, whereas 5R and 5S, having α-amino tailed asymmetric double heads, showed poor potency. Compounds 6−8, having an α-aminomethyl tail with two aminopyridine head groups, showed improved potency by 1- to 5-fold. The orientation of the nitrogen on the middle aromatic ring was important; compound 7 was a relatively poor inhibitor of nNOS when the nitrogen was located in the narrow arc of the middle ring. Although the α-aminomethyl derivatives 6, 7, and 8 did not display improved binding affinity for nNOS compared with that of their parent molecules 31, 32, and 2, it is noteworthy that all of those α-aminomethyl derivatives have better selectivity (Table 1) against iNOS and eNOS. In general, the α-aminomethyl derivatives are 2- to 3-fold (i/n) and 3- to 10-fold (e/n) more selective than their parent molecules. To further explore the inhibition potency and selectivity of racemic 8, each enantiomer was prepared and assayed with the three NOS isoforms. Compound 8R, the (R)enantiomer of 8, showed excellent potency (Ki = 24 nM) for nNOS with a 273-fold selectivity over iNOS and a 2822-fold selectivity over eNOS, the best in the series. We also assayed compound 8 against human nNOS to explore whether the interactions are acceptable for the human isoform as well. Human nNOS is very similar to rat nNOS, except that the hydrophobic pocket surrounded by Met341, Leu337, and Tyr706 in rat nNOS is replaced by Met340, His342, and Tyr711. This pocket is where the second headgroup of this series of double-headed inhibitors fits. The inhibitory potency of 8 for human nNOS is 90 nM, similar to that (70 nM) of compound 1,25 which is our most potent human nNOS inhibitor with potential selectivity over human iNOS and eNOS. Structure−Activity Relationship Studies. To aid in the structure−activity relationship studies for the series of doubleheaded aminopyridine compounds, we first determined the crystal structures of nNOS in complex with 3S, 3R, 4, and 5 that bear an α-amino group (Table 1). Both 3S and 3R are able to bind to nNOS with both aminopyridine heads involved in Hbonds, one with Glu592 and the other with the propionate of the heme D-ring, respectively (Figure 3). In contrast, the parent compound (31) showed only one aminopyridine H-bonded with Glu592, while the rest of the inhibitor was badly disordered in structure (unpublished data). Introducing an αamino group next to the center phenyl ring in 3S and 3R helps to stabilize the double-headed binding. Interestingly, the binding orientation of 3 is dependent on the chirality leading to different positions for the α-amine. The α-amino group of 3S is next to the aminopyridine that H-bonds with Glu592 and points downward to the heme (Figure 3A), while the α-amino group of 3R is on the side of the aminopyridine that H-bonds with the propionate of the heme D-ring (Figure 3B). This amino group can H-bond with a conserved water molecule that is bridging between the H4B and the propionate of heme A-ring rather than replace the water as expected from modeling. When the α-amino group is replaced with a hydroxyl group in 4, the compound can no longer achieve a double-headed binding (Figure 3C). Instead, the aminopyridine next to the 4388



Article

Figure 4. Active site structures of bovine eNOS in complex with (A) 3R (PDB: 4CTY) and (B) 5 or 5S (PDB: 4CTZ). The Fo − Fc omit electron density for the inhibitor is displayed at the 2.5 σ contour level. The major H-bonds are depicted with dashed lines.

eNOS explains the 363-fold selectivity observed for 5S (Table 1). The double-headed compounds with only an α-amino group on the side chain (3S, 3R, 4, and 5S) do not quite reach the position of the pyrroline amine of parent compound 1 that can replace a water molecule (Figure 1B). This might be why the low nanomolar affinity and good isoform selectivity of 1 had not been achieved with 3−5. We, therefore, introduced an aminomethyl group to racemic compounds 6, 7, and 8 (Figure 2). We also designed three different center aromatic rings to further explore the influence of the polarity of this ring to the potency and selectivity of inhibitors. The inhibitory assays indicated that 6 and 8 indeed have improved potency (Table 1). The crystal structures of 6 and 8 bound to nNOS showed that they share a similar double-headed binding mode (Figure 5). Although the racemic samples were used for crystal preparation, the resulting structures were dominated by the R-enantiomer in both cases. The structure of nNOS-6R (Figure 5A) overlays well with that of nNOS-3R (Figure 3B) except that the aminomethyl group of 6 replaces a water molecule, thereby allowing it to make H-bonds with both the H4B and the propionate of heme A-ring, as was expected from the design. The better interactions from the aminomethyl group of 6 to both the H4B and the propionate afford a 2-fold improvement in potency compared with that of 3 (Table 1). Compound 8 exhibits even better potency with nNOS than does 6 (Table 1), which may result from the extra nitrogen on the center pyridine ring of 8 (Figure 5B) versus the phenyl ring in 6 (Figure 5A). This pyridine ring in 8 is pulled a bit further up toward Asp597 in nNOS than is the phenyl ring of 6, making water bridged H-bonds with the acidic residue and another weak H-bond (3.4 Å) with Gln478 (Figure 5B). We have also designed compound 7, having the pyridine nitrogen

Figure 5. Active site structures of rat nNOS in complex with (A) 6 (PDB: 4CTU), (B) 8 or 8R (PDB: 4CTW), and (C) 7 or 7S (PDB: 4CTV). The Fo − Fc omit electron density for the inhibitor is displayed at the 2.5 σ contour level. The major H-bonds are depicted with dashed lines. The weak H-bond in nNOS-8R is labeled with a distance in Å.

at an ortho-position relative to the other two substituents in contrast to the meta-position in 8. To our surprise, 7 binds to nNOS (Figure 5C) in an orientation that is 180° flipped from that of 8 (Figure 5B). Moreover, it is the S-enantiomer of 7 that populates the structure, even though a racemic mixture of 7 was used in crystal preparation. In this binding orientation, the aminomethyl group of 7 makes a tight H-bond with Glu592. The positions of the neighboring aminopyridine, H-bonding to Glu592, and the center phenyl ring, bending toward the heme, overlay well with their counterparts in 5S (Figure 3D). In addition, the second aminopyridine of 7 can also reach the site of the heme D-ring propionate in a double-headed mode (Figure 5C). Although in this flipped binding orientation, 7 shows poorer potency than 6 or 8, it does have an impressive isoform selectivity (Table 1). We have tried, but failed, to get 4389



Article

an eNOS-7 structure because the inhibitor was badly disordered except for the aminopyridine that H-bonds with Glu363 (data not shown). From the known structure of eNOS5S (Figure 4B), we reasoned that the poor binding affinity of 7 toward eNOS must result from its inability to establish direct H-bonds with Glu363 by its aminomethyl group because of the more electropositive environment of Glu363 and Asn368 in eNOS, which would not allow a positively charged aminomethyl group to enter the pocket. Compound 8 exhibits the best potency toward nNOS in the series, which prompted us to further explore the effects of chirality on the inhibitory potency and selectivity. Enantiopure 8S and 8R were synthesized, and crystal structures of both enantiomers bound to nNOS and eNOS were determined. Indeed, 8R gives the best potency (24 nM) with nNOS and 2822-fold selectivity for nNOS over eNOS (Table 2). Both 8S and 8R show very similar double-headed binding to nNOS; the nNOS-8R structure is the same as that of nNOS-8 (Figure 5B), while the nNOS-8S structure (Figure 6A) can be, more or less, superimposed on nNOS-8R, with the only exception being around the chiral center. Also, the center pyridine of 8S can make a better H-bond (2.9−3.0 Å) with Gln478. Regardless of the chirality, the aminomethyl group in both 8S and 8R can replace a water molecule equally well, thus making direct Hbonds with both the H4B and the propionate of the heme Dring. The similarity in structure seems to be reasonable considering the ∼3-fold difference in potency between 8S and 8R. The eNOS-8S structure (Figure 6B) shows a one-headed binding mode since the access to the heme D-ring propionate from the second aminopyridine of 8S is blocked by the Tyr477 side chain, which remains in its position, H-bonding with the same propionate. Therefore, the second aminopyridine group is poorly defined in structure, although the first aminopyridine and aminomethyl groups maintain the H-bonding interactions that are also observed in nNOS-8S (Figure 6A). To our surprise, we found no significant difference in the binding mode of 8R to eNOS (Figure 6C) versus that to nNOS (Figure 5B). With a double-headed binding mode, the center pyridine of 8R in eNOS points toward the direction of Asn368, just like its counterpart in nNOS, which points toward Asp597. In contrast, the central pyridine ring of 8S in eNOS, in a one-headed binding, bends away from Asn368 and makes a better H-bond with Gln249 (Figure 6B). This may be part of the reason that 8S binds better to eNOS than does 8R (Table 2). The structural binding mode alone cannot explain the 2822fold isoform selectivity of 8R for nNOS over eNOS. This is not the first time that a highly selective NOS inhibitor shows an almost identical binding mode to nNOS and eNOS. The previous examples were parent compound 122 and its aminoanalogue.24 To better understand the basis for isoform selectivity, we turned to a computational approach that proved useful in previous studies with NOS inhibitors.24 Using the MM-PBSA approach, we first calculated the ΔG for 8R bound to nNOS. The value from this calculation is called PBtotal and includes the solvation and enthalpic terms but not the decrease in entropy upon inhibitor moving from solution to the active site. Since we are examining exactly the same inhibitor binding to nNOS and eNOS, ignoring this entropic contribution introduces little error. Table 3 provides the results of the MMPBSA calculations. This approach comes to within 0.8 kcal/mol of estimating the ΔG of binding to eNOS and also shows that the primary difference between eNOS and nNOS is the more favorable

Figure 6. Active site structures of rat nNOS in complex with (A) 8S (PDB: 4CTX) and of bovine eNOS with (B) 8S (PDB: 4CU1) and (C) 8R (PDB: 4CU0). The Fo − Fc omit electron density for inhibitor is displayed at 2.5 σ contour level. The major H-bonds are depicted with dashed lines. The weak H-bond in eNOS-8R is labeled with a distance in Å.

Table 3. Results of the MM-PBSA Calculations (kcal/mol) protein

ELE

VDW

PBele

PBtotal

ΔGcalc

ΔGexp

nNOS eNOS

−1018.7 −818.7

−44.92 −41.82

−41.1 −9.57

−92.2 −57.51

−10.4a −6.5

−10.4 −5.7

ΔGcalc for nNOS was normalized to match ΔGexp by dividing PBtotal by ΔGexp and then using this conversion factor to compute ΔGclac for eNOS. ELE and VDW are the electrostatic and van der Waals contributions, respectively, to inhibitor binding. PBele is the sum of ELE and the reaction field energy calculated with the Poisson− Boltzmann equation and thus represents the total electrostatic component of inhibitor binding. PBtotal is the sum of all energies and represents the total free energy of binding uncorrected for the entropic contribution of the inhibitor moving from solution to the active site. a

4390



Article

the compounds was evaluated on an Beckman Gold reverse phase analytical HPLC system using an Phenonemex Gemini C-18 (4.6 × 250 mm, 5 μm) or Luna C-8 (4.6 × 250 mm, 5 μm) reverse phase columns with UV absorbance and evaporative light scattering detection. Purities of all compounds that were subjected to the biological assay were >95%. The direct chiral resolutions of racemic 8R and 8S were performed on a Beckman Gold HPLC system using a Chiralcel OD-H HPLC column (Daicel, 250 × 4.6 mm i.d., 5 mm). Hexanes and isopropanol (85−92% hexanes) were used as the mobile phases. The operation temperature was 25 °C, and the flow rate was 0.8 mL/min with 254 nm UV detection. Optical rotations were measured on a PerkinElmer Model 341 digital readout polarimeter. General Procedure for the Deprotection of 2-(2,5-Dimethyl1H-pyrrol-1-yl)pyridine Derivatives Using Microwave Irradiation:27 Method A. To a 5 mL microwave vial equipped with a magnetic stir bar was added the protected aminopyridine (1.0 mmol), ethanol (2.5 mL), and concentrated hydrochloric acid (0.5 mL). After being capped, the vial was shaken vigorously and then heated in the microwave irradiator for 20 min at 120 °C (as recorded via the IR sensor of the microwave instrument). After being cooled to room temperature, the reaction mixture was concentrated in vacuo and purified by flash column chromatography using a SiliaSep C18 flash cartridge (25g, 40−63 μm/230−400 mesh, pore size 60 Å) with 5 to 80% MeOH in water as the mobile phase. (S)-6-(2-Amino-2-(3-(2-(6-amino-4-methylpyridin-2-yl)ethyl)phenyl)ethyl)-4-methylpyridin-2-amine (3S) and (R)-6-(2-Amino-2(3-(2-(6-amino-4-methylpyridin-2-yl)ethyl)phenyl)ethyl)-4-methylpyridin-2-amine (3R). The title compounds were synthesized using general method A. 3S (55 mg, 58%) was prepared as a pale yellow gel from 13a (180 mg, 0.26 mmol), and 3R (50 mg, 60%) was prepared as a pale yellow gel from 13b (160 mg, 0.23 mmol). 3S, [α]20D = +72.1° (c 3 g/L, MeOH); 3R, [α]20D = −79.2° (c 2 g/L, MeOH). 1H NMR (500 MHz, MeOD) δ 7.22−7.17 (m, 1H), 7.14 (qt, J = 2.9, 1.9, 1.4 Hz, 2H), 7.05 (dt, J = 7.4, 1.6 Hz, 1H), 6.25 (s, 1H), 6.24 (ss, 2H), 6.18 (s, 1H), 4.21 (t, J = 7.1 Hz, 1H), 2.89 (dd, J = 9.6, 6.4 Hz, 2H), 2.81 (d, J = 7.1 Hz, 2H), 2.79−2.71 (m, 2H), 2.13 (s, 3H), 2.11 (s, 3H); 13C NMR (126 MHz, MeOD) δ 160.86, 160.70, 159.73, 157.35, 151.03, 150.86, 145.85, 143.10, 129.46, 128.35, 127.82, 125.18, 115.89, 114.67, 108.22, 107.92, 57.15, 48.09, 40.64, 37.32, 21.08, 21.05. HRMS (ESI): calcd for C22H28N5 [M + H]+, 362.2339; found, 362.2338. (R)-6-(2-Amino-2-(3-(2-(4-methylpyridin-2-yl)ethyl)phenyl)ethyl)4-methylpyridin-2-amine (5R) and (S)-6-(2-Amino-2-(3-(2-(4-methylpyridin-2-yl)ethyl)phenyl)ethyl)-4-methylpyridin-2-amine (5S). Method A was used to prepare 5R (153 mg, 65%, colorless gel) from 19a (360 mg, 0.682 mmol) and 5S (22 mg, 60%, colorless gel) from 19b (55 mg, 0.104 mmol). 5S, [α]20D = −20.0° (c 2 g/L, MeOH); 5R, [α]20D = +19.2° (c 2 g/L, MeOH). 1H NMR (500 MHz, MeOD) δ 8.54 (d, J = 6.0 Hz, 1H), 7.79 (s, 1H), 7.68 (d, J = 5.9 Hz, 1H), 7.63 (s, 1H), 7.51−7.37 (m, 2H), 7.36−7.29 (m, 1H), 6.68 (s, 2H), 4.87 (d, J = 7.8 Hz, 1H), 3.51 (dd, J = 14.7, 7.9 Hz, 1H), 3.42− 3.34 (m, 3H), 3.21−3.14 (m, 2H), 2.62 (s, 3H), 2.34 (s, 3H); 13C NMR (126 MHz, MeOD) δ 162.50, 158.89, 156.51, 156.12, 143.95, 142.17, 141.24, 137.24, 131.02, 130.88, 129.16, 128.68, 127.00, 126.98, 116.76, 112.22, 55.32, 38.40, 35.76, 22.36, 21.96. HRMS (ESI): calcd for C22H27N4 [M + H]+, 347.2230; found, 347.2229. 6-(3-Amino-2-(3-(2-(6-amino-4-methylpyridin-2-yl)ethyl)phenyl)propyl)-4-methylpyridin-2-amine (6). Method A was used to prepare 6 (144 mg, 43%, pale yellow gel) from 24 (475 mg, 0.89 mmol). 1H NMR (500 MHz, MeOD) δ 7.43 (s, 1H), 7.32 (t, J = 7.6 Hz, 1H), 7.21 (dt, J = 7.5, 1.8 Hz, 2H), 6.68 (s, 2H), 6.62 (s, 1H), 6.50 (s, 1H), 3.60−3.51 (m, 1H), 3.40−3.34 (m, 2H), 3.29 (dd, J = 14.3, 6.1 Hz, 1H), 3.07−3.00 (m, 5H), 2.38 (s, 3H), 2.28 (s, 3H); 13C NMR (126 MHz, MeOD) δ 159.08, 158.70, 155.78, 150.00, 147.22, 142.47, 139.40, 130.61, 129.59, 129.31, 127.58, 116.07, 114.88, 111.16, 110.83, 44.87, 44.69, 37.99, 35.93, 35.61, 22.01, 21.91. HRMS (ESI): calcd for C23H30N5 [M + H]+, 376.2496; found, 376.2502. 6-(3-Amino-2-(6-(2-(6-amino-4-methylpyridin-2-yl)ethyl)pyridin2-yl)propyl)-4-methylpyridin-2-amine (7). Method A was used to prepare 7 (87 mg, two step 43%, pale yellow gel) from crude 30a (300 mg). 1H NMR (500 MHz, MeOD) δ 7.83 (m, 1H), 7.38 (m, 1H), 7.33 (m, 1H), 6.73 (s, 1H), 6.70 (s, 1H), 6.65 (s, 1H), 6.45 (s, 1H),

electrostatic interactions between the inhibitor and nNOS than those in eNOS. We attribute this primarily to Asp597 in nNOS vs Asn368 in eNOS, even though Asp597 does not directly contact the inhibitor. If the active site pocket exhibits a relatively low dielectric milieu, then long-range electrostatic stabilization by Asp597 on both the aminomethyl group and the central pyridine of 8R could be quite substantial. However, when the binding of 8R to eNOS is forced into the same double-headed mode as that with nNOS, the penalty toward electrostatic interactions leads to poorer binding and, therefore, higher isoform selectivity.

■

CONCLUSIONS A series of α-amino functionalized aminopyridine derivatives (3−8) were designed based on the comparison and rationale that the pyrrolidine of lead 1, interacting with a heme propionate and H4B, is the key to isoform selectivity. Therefore, an α-amino or aminomethyl group has been installed in the other lead (2) to probe the structure−selectivity relationship, while providing a substantially simple scaffold that retains selectivity (Figure 1). In general, the symmetric doubleheaded aminopyridine compounds (3 and 4) showed better potency than the asymmetric one (5) because the former can establish H-bonds through both head groups. Although a simple α-amino group installed on the side chain can stabilize the double-headed inhibitor binding mode, only the aminomethyl group is long enough to reach a water site. By replacing the water molecule, the inhibitor is able to make H-bonds with both the H4B and the propionate of the heme A-ring. These interactions seem to be crucial to gain the isoform selectivity for the inhibitors. Compound 8R, the best inhibitor discovered in this study, exhibits excellent nNOS potency (24 nM) and isoform selectivity (273-fold for i/n and 2822-fold for e/n). It also showed

Nitric Oxide Synthase Inhibitors That Interact with Both Heme

Recommend Documents